Jensen Huang has articulated one of the most uncomfortable consequences of the ongoing tech war between the United States and China. According to Nvidia’s CEO, the company’s market share in AI accelerators in China has dropped to 0%. This isn’t just another dip in a challenging market; it’s the disappearance, at least in terms of direct sales, of a company that until recently clearly dominated one of the world’s largest tech markets.
The statement carries political, industrial, and economic weight. Huang argues that U.S. export restrictions on advanced chips “have largely failed” because they’ve opened up a huge market to local competitors. His reasoning is straightforward: if U.S. companies can’t sell in China, China continues building its own AI data centers, sourcing alternative solutions, accelerating its native chips, and pushing developers to rely less on Nvidia’s ecosystem.
The ban hasn’t stopped Chinese AI, it has changed its providers
Washington’s strategy originated from a national security perspective. The U.S. aims to prevent China from accessing chips capable of training advanced models, operating supercomputers, or accelerating military applications. Major restrictions first appeared in October 2022 and expanded in 2023, targeting certain advanced computing chips, supercomputing systems, and semiconductor manufacturing equipment.
However, the market doesn’t freeze when a door closes. Nvidia has adapted its products to continue selling in China within permissible limits, as with the H20 family. But that route also faced limitations. In April 2025, Nvidia was informed that a license was required to export H20 to China. The immediate impact was notable: a $4.5 billion inventory and purchase commitments charge, $4.6 billion in H20 sales recognized before the new restrictions, and an additional $2.5 billion in potential revenues delayed during Nvidia’s first fiscal quarter of 2026.
The situation worsened in the following quarter. Nvidia reported no H20 sales to Chinese-based clients during Q2 of FY2026, and forecasts did not include H20 shipments to China. This indicates Nvidia has experienced an abrupt hit and is managing its business as if China no longer plays a key role in its AI accelerators market, at least in the immediate term.
This is where Huang’s firm stance makes sense. From Nvidia’s viewpoint, restrictions haven’t eliminated Chinese demand for AI computing—they have forced local customers to seek substitutes. Meanwhile, Chinese players like Huawei, Cambricon, Moore Threads, MetaX, and others have gained ground in a market that once almost automatically leaned toward Nvidia.
China accelerates its national alternative
The key question is whether China can truly replace Nvidia. In raw performance, software, developer tools, interconnection, and ecosystem maturity, Nvidia still holds a significant edge. CUDA isn’t just a software layer; it’s a network of libraries, optimizations, frameworks, documentation, and accumulated expertise built over years. Replicating that isn’t achievable overnight.
However, China doesn’t need to match Nvidia in all areas to reduce dependence. It can begin by focusing on inference, more controlled deployments, internal loads for large platforms, and models optimized for local hardware. That pathway is already underway. Sector reports point to rapid growth among Chinese providers, with Huawei as a candidate to take an increasing share of the domestic market and Cambricon gaining strong traction among large clients.
Bernstein previously estimated Nvidia’s share of the Chinese AI GPU market could fall from 66% in 2024 to around 8% in the coming years, with domestic providers covering a growing portion of demand. Huang now states that for Nvidia, direct sales in China have already dropped to zero, though this figure warrants nuance: it doesn’t necessarily mean Nvidia hardware isn’t installed, imported through third parties, or used in legacy environments.
This nuance is important. A 0% figure is a powerful commercial and political headline but doesn’t fully depict the actual use of Nvidia chips inside China. Still, the direction is clear: the more Nvidia is out of the market, the greater the incentives for Chinese companies to adapt models, frameworks, and data centers to local alternatives. As their ecosystem matures, it will become harder for Nvidia to regain market share if restrictions ease.
The chips policy enters a more uncomfortable phase
The debate lacks a simple answer. Advocates of bans argue that allowing massive sales of advanced accelerators to China could bolster the strategic military, surveillance, and supercomputing capabilities of a rival. This is a real concern, which is why export controls can’t be based solely on business considerations.
Huang’s critique points elsewhere: if policies succeed in ousting U.S. companies without sufficiently slowing Chinese progress, the outcome could be worse for Washington. The U.S. would lose revenue, technological influence, and presence in a huge market, while China would accelerate development of its own supply chain with increased urgency and political backing.
The evolving rules themselves reflect this tension. In January 2026, the Commerce Department revised its export licensing policy, analyzing case-by-case approvals for chips like Nvidia’s H200 or AMD’s MI325X to China, always under certain security conditions. This flexibility indicates an awareness within Washington that an outright ban has strategic costs.
Meanwhile, Nvidia continues to grow strongly outside China, driven by global demand for Blackwell, large AI clusters, and hyperscaler investments. But China remains too significant to ignore. Losing that market wouldn’t sink Nvidia short-term, but it could shift the industry’s long-term balance. The company grants space to local rivals, diminishes its ability to set standards in China, and inadvertently fuels the very technological independence the U.S. aims to contain.
The Nvidia case exemplifies a paradox of the new tech geopolitical landscape. The most advanced chips are so strategic that governments want to control them, yet overreaching can push other countries to develop alternatives. Sometimes these efforts succeed quickly; sometimes they take time. But when a market as large as China feels compelled to replace a technology, investment inevitably follows.
Huang’s statement shouldn’t be viewed solely as a complaint from a CEO losing sales. It’s also a warning about industrial policy design: restricting a competitor doesn’t suffice if the move accelerates its independence. In the AI race, the U.S. still holds enormous advantages in chips, software, cloud infrastructure, and talent. But abandoning China altogether could undermine one of its most powerful assets: shaping global technology development based on U.S. innovation.
Frequently Asked Questions
What did Jensen Huang say about Nvidia in China?
He stated that Nvidia’s market share in China’s AI accelerators has dropped to 0%, and that U.S. export restrictions have had a counterproductive effect.
Why does the U.S. limit AI chip sales to China?
Washington cites national security reasons. Controls aim to prevent China from using advanced chips for supercomputing, military applications, surveillance, or strategic AI development.
Does the 0% figure mean Nvidia chips aren’t in China?
Not necessarily. It refers to Nvidia’s direct commercial presence in the Chinese AI accelerators market. Hardware could still be present via installations, imports through third parties, or legacy systems.
Who might fill Nvidia’s gap in China?
Chinese providers like Huawei, Cambricon, Moore Threads, MetaX, and others are trying to cover some of the demand. Nvidia maintains a clear lead in performance and software, but China is rapidly evolving its own supply chain.